SFs (Sub Functions) currently use IRQs from the global IRQ table their
parent Physical Function have. In order to better scale, we need to
allocate more IRQs and share them between different SFs.

Driver will maintain 3 separated irq pools:
1. A pool that serve the PF consumer (PF's netdev, rdma stacks), similar
to what the driver had before this patch. i.e, this pool will share irqs
between rdma and netev, and will keep the irq indexes and allocation
order. The last is important for PF netdev rmap (aRFS).

2. A pool of control IRQs for SFs. The size of this pool is the number
of SFs that can be created divided by SFS_PER_IRQ. This pool will serve
the control path EQs of the SFs.

3. A pool of completion data path IRQs for SFs transport queues. The
size of this pool is:
num_irqs_allocated - pf_pool_size - sf_ctrl_pool_size.
This pool will served netdev and rdma stacks. Moreover, rmap is not
supported on SFs.

Sharing methodology of the SFs pools is explained in the next patch.

Important note: rmap is not supported on SFs because rmap mapping cannot
function correctly for IRQs that are shared for different core/netdev RX
rings.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Store newly created IRQs in the xarray DB instead of a static array,
so we will be able to store only IRQs which are being used.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

IRQs are being simplified in order to ease their sharing and any feature
specific object will be moved to upper layer.
Hence we move rmap object into eq_table.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Extend mlx5_irq_request so that IRQs will be requested upon EQ creation,
and not on driver boot.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

In next patches, IRQs will be requested according to demand, instead of
statically on driver boot.
Also, currently, rmap is managed by the IRQ layer. rmap management will
move out from the IRQ layer in future patches.

Therefore, we want to remove the IRQ from the rmap, when IRQ is destroyed,
instead of removing all the IRQs from the rmap when irq_table is destroyed.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

The eq.[c|h] files are under major rewrite. so use this opportunity and
update their copyright and license texts.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

The users of EQ are running their code on different CPUs and with
various affinity patterns. Move the cpumask setting close to their
actual usage.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Introduce new API that will allow IRQs users to hold a pointer to
mlx5_irq.
In the end of this series, IRQs will be allocated on demand. Hence,
this will allow us to properly manage and use IRQs.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Shared IRQ are consumed by multiple EQ users and in order to properly
initialize and later release such IRQs, we add kref counting of IRQ
structure.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Lag is used to combine two PCI functions of the same HCA into a single
logical unit. This is a core functionality and as such should be managed by
the core driver. Currently this isn't the case. While we store the lag
software structure inside the lower device, its lifetime (creation /
destruction) is dictated by the mlx5e part. Change the ownership model so
lag is tied to the lifetime of the lower level driver instead to the
mlx5e part.
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

If MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV is set it means the device is going
down and mlx5_rescan_drivers_locked() shouldn't be called.
With this patch and the previous one in the series, unbinding a PCI
function when its netdev is part of a bond works and leaves the system in a
working state.
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

When a net device is removed (can happen if the PCI function is unbound
from the system) it's not enough to destroy the hardware lag. The system
should recreate the original devices that were present before the lag.
As the same flow is done when a net device is removed from the bond
refactor and reuse the code.
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

There is not strong reason to have both WWAN and WWAN_CORE symbols,
Let's build the WWAN core framework when WWAN is selected, in the
same way as for other subsystems.

This fixes issue with mhi_net selecting WWAN_CORE without WWAN and
reported by kernel test robot:

Kconfig warnings: (for reference only)
   WARNING: unmet direct dependencies detected for WWAN_CORE
   Depends on NETDEVICES && WWAN
   Selected by
   - MHI_NET && NETDEVICES && NET_CORE && MHI_BUS

Fixes: 9a44c1cc ("net: Add a WWAN subsystem")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

After the blamed patch, __skb_flow_dissect() on the DSA master stopped
adjusting for the length of the DSA headers. This is because it was told
to adjust only if the needed_headroom is zero, aka if there is no DSA
header. Of course, the adjustment should be done only if there _is_ a
DSA header.

Modify the comment too so it is clearer.

Fixes: 4e500251 ("net: dsa: generalize overhead for taggers that use both headers and trailers")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The struct sja1105_regs tables are not modified during the runtime of
the driver, so they can be made constant. In fact, struct sja1105_info
already holds a const pointer to these.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Vladimir Oltean says:

====================
Fixes and improvements to TJA1103 PHY driver

This series contains:
- an erratum workaround for the TJA1103 PHY integrated in SJA1110
- an adaptation of the driver so it prints less unnecessary information
  when probing on SJA1110
- a PTP RX timestamping bug fix and a clarification patch

Targeting net-next since the PHY support is currently in net-next only.

Changes in v3:
Added one more patch which improves the readability of
nxp_c45_reconstruct_ts.

Changes in v2:
Added a comment to the hardware workaround procedure.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The SJA1110 switch integrates TJA1103 PHYs, but in SJA1110 switch rev B
silicon, there is a bug in that the registers for selecting the 100base-T1
autoneg master/slave roles are not writable.

To enable write access to the master/slave registers, these additional
PHY writes are necessary during initialization.

The issue has been corrected in later SJA1110 silicon versions and is
not present in the standalone PHY variants, but applying the workaround
unconditionally in the driver should not do any harm.
Suggested-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

The reconstruction procedure for partial timestamps reads the current
PTP time and fills in the low 2 bits of the second portion, as well as
the nanoseconds portion, from the actual hardware packet timestamp.
Critically, the reconstruction procedure works because it assumes that
the current PTP time is strictly larger than the hardware timestamp was:
it detects a 2-bit wraparound of the 'seconds' portion by checking whether
the 'seconds' portion of the partial hardware timestamp is larger than
the 'seconds' portion of the current time. That can only happen if the
hardware timestamp was captured by the PHY during the last phase of a
'modulo 4 seconds' interval, and the current PTP time was read by the
driver during the initial phase of the next 'modulo 4 seconds' interval.

The partial RX timestamps are added to priv->rx_queue in
nxp_c45_rxtstamp() and they are processed potentially in parallel by the
aux worker thread in nxp_c45_do_aux_work(). This means that it is
possible for nxp_c45_do_aux_work() to process more than one RX timestamp
during the same schedule.

There is one premature optimization that will cause issues: for RX
timestamping, the driver reads the current time only once, and it uses
that to reconstruct all PTP RX timestamps in the queue. For the second
and later timestamps, this will be an issue if we are processing two RX
timestamps which are to the left and to the right, respectively, of a
4-bit wraparound of the 'seconds' portion of the PTP time, and the
current PTP time is also pre-wraparound.

 0.000000000        4.000000000        8.000000000        12.000000000
 |..................|..................|..................|............>
                 ^ ^ ^ ^                                            time
                 | | | |
                 | | | process hwts 1 and hwts 2
                 | | |
                 | | hwts 2
                 | |
                 | read current PTP time
                 |
                 hwts 1

What will happen in that case is that hwts 2 (post-wraparound) will use
a stale current PTP time that is pre-wraparound.
But nxp_c45_reconstruct_ts will not detect this condition, because it is
not coded up for it, so it will reconstruct hwts 2 with a current time
from the previous 4 second interval (i.e. 0.something instead of
4.something).

This is solvable by making sure that the full 64-bit current time is
always read after the PHY has taken the partial RX timestamp. We do this
by reading the current PTP time for every timestamp in the RX queue.

Fixes: 514def5d ("phy: nxp-c45-tja11xx: add timestamping support")
Cc: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

nxp_c45_reconstruct_ts() takes a partial hardware timestamp in @hwts,
with 2 bits of the 'seconds' portion, and a full PTP time in @ts.

It patches in the lower bits of @hwts into @ts, and to ensure that the
reconstructed timestamp is correct, it checks whether the lower 2 bits
of @hwts are not in fact higher than the lower 2 bits of @ts. This is
not logically possible because, according to the calling convention, @ts
was collected later in time than @hwts, but due to two's complement
arithmetic it can actually happen, because the current PTP time might
have wrapped around between when @hwts was collected and when @ts was,
yielding the lower 2 bits of @ts smaller than those of @hwts.

To correct for that situation which is expected to happen under normal
conditions, the driver subtracts exactly one wraparound interval from
the reconstructed timestamp, since the upper bits of that need to
correspond to what the upper bits of @hwts were, not to what the upper
bits of @ts were.

Readers might be confused because the driver denotes the amount of bits
that the partial hardware timestamp has to offer as TS_SEC_MASK
(timestamp mask for seconds). But it subtracts a seemingly unrelated
BIT(2), which is in fact more subtle: if the hardware timestamp provides
2 bits of partial 'seconds' timestamp, then the wraparound interval is
2^2 == BIT(2).

But nonetheless, it is better to express the wraparound interval in
terms of a definition we already have, so replace BIT(2) with
1 + GENMASK(1, 0) which produces the same result but is clearer.
Suggested-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Cc: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

The SJA1110 switch integrates these PHYs, and they do not have support
for timestamping. This message becomes quite overwhelming:

[   10.056596] NXP C45 TJA1103 spi1.0-base-t1:01: the phy does not support PTP
[   10.112625] NXP C45 TJA1103 spi1.0-base-t1:02: the phy does not support PTP
[   10.167461] NXP C45 TJA1103 spi1.0-base-t1:03: the phy does not support PTP
[   10.223510] NXP C45 TJA1103 spi1.0-base-t1:04: the phy does not support PTP
[   10.278239] NXP C45 TJA1103 spi1.0-base-t1:05: the phy does not support PTP
[   10.332663] NXP C45 TJA1103 spi1.0-base-t1:06: the phy does not support PTP
[   15.390828] NXP C45 TJA1103 spi1.2-base-t1:01: the phy does not support PTP
[   15.445224] NXP C45 TJA1103 spi1.2-base-t1:02: the phy does not support PTP
[   15.499673] NXP C45 TJA1103 spi1.2-base-t1:03: the phy does not support PTP
[   15.554074] NXP C45 TJA1103 spi1.2-base-t1:04: the phy does not support PTP
[   15.608516] NXP C45 TJA1103 spi1.2-base-t1:05: the phy does not support PTP
[   15.662996] NXP C45 TJA1103 spi1.2-base-t1:06: the phy does not support PTP

So reduce its log level to debug.

Cc: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Zhou Yanjie says:

====================
Add Ingenic SoCs MAC support.

v2->v3:
1.Add "ingenic,mac.yaml" for Ingenic SoCs.
2.Change tx clk delay and rx clk delay from hardware value to ps.
3.return -EINVAL when a unsupported value is encountered when
  parsing the binding.
4.Simplify the code of the RGMII part of X2000 SoC according to
  Andrew Lunn’s suggestion.
5.Follow the example of "dwmac-mediatek.c" to improve the code
  that handles delays according to Andrew Lunn’s suggestion.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for Ingenic SoC MAC glue layer support for the stmmac
device driver. This driver is used on for the MAC ethernet controller
found in the JZ4775 SoC, the X1000 SoC, the X1600 SoC, the X1830 SoC,
and the X2000 SoC.
Signed-off-by: 周琰杰 (Zhou Yanjie) <zhouyanjie@wanyeetech.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add the dwmac bindings for the JZ4775 SoC, the X1000 SoC,
the X1600 SoC, the X1830 SoC and the X2000 SoC from Ingenic.
Signed-off-by: 周琰杰 (Zhou Yanjie) <zhouyanjie@wanyeetech.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Oleksandr Mazur says:

====================
Marvell Prestera driver implementation of devlink functionality.

This patch series implement Prestera Switchdev driver devlink traps,
that are registered within the driver, as well as extend current devlink
functionality by adding new hard drop statistics counter, that could be
retrieved on-demand: the counter shows number of packets that have been
dropped by the underlying device and haven't been passed to the devlink
subsystem.

The core prestera-devlink functionality is implemented in the prestera_devlink.c.

The patch series also extends the existing devlink kernel API:
 - devlink: add trap_drop_counter_get callback for driver to register - make it possible
   to keep track of how many packets have been dropped (hard) by the switch device, before
   the packets even made it to the devlink subsystem (e.g. dropped due to RXDMA buffer
   overflow).

The core features that extend current functionality of prestera Switchdev driver:
 - add logic for driver traps and drops registration (also traps with DROP action).
 - add documentation for prestera driver traps and drops group.

PATCH v2:
 1) Rebase whole series on top of latest mater;
 2) Remove storm control-related patches, as they're out of devlink
    scope;
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Add documentation for the devlink feature prestera switchdev driver supports:
add description for the support of the driver-specific devlink traps
(include both traps with action TRAP and action DROP);
Signed-off-by: Oleksandr Mazur <oleksandr.mazur@plvision.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add traps that have init_action being set to DROP.
Add 'trap_drop_counter_get' (devlink API) callback implementation,
that is used to get number of packets that have been dropped by the HW
(traps with action 'DROP').
Add new FW command CPU_CODE_COUNTERS_GET.
Signed-off-by: Oleksandr Mazur <oleksandr.mazur@plvision.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add devlink traps registration (with corresponding groups) for
all the traffic types that driver traps to the CPU;
prestera_rxtx: report each packet trapped to the CPU (RX) to the
prestera_devlink;
Signed-off-by: Oleksandr Mazur <oleksandr.mazur@plvision.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add hard drop counter check testcase, to make sure netdevsim driver
properly handles the devlink hard drop counters get/set callbacks.
Signed-off-by: Oleksandr Mazur <oleksandr.mazur@plvision.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

Whenever query statistics is issued for trap with DROP action,
devlink subsystem would also fill-in statistics 'dropped' field.
In case if device driver did't register callback for hard drop
statistics querying, 'dropped' field will be omitted and not filled.
Add trap_drop_counter_get callback implementation to the netdevsim.
Add new test cases for netdevsim, to test both the callback
functionality, as well as drop statistics alteration check.
Signed-off-by: Oleksandr Mazur <oleksandr.mazur@plvision.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

testing: selftests: net: forwarding: add devlink-required functionality to test (hard) dropped stats field

Add devlink_trap_drop_packets_get function, as well as test that are
used to verify devlink (hard) dropped stats functionality works.
Signed-off-by: Oleksandr Mazur <oleksandr.mazur@plvision.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

Whenever query statistics is issued for trap, devlink subsystem
would also fill-in statistics 'dropped' field. This field indicates
the number of packets HW dropped and failed to report to the device driver,
and thus - to the devlink subsystem itself.
In case if device driver didn't register callback for hard drop
statistics querying, 'dropped' field will be omitted and not filled.
Signed-off-by: Oleksandr Mazur <oleksandr.mazur@plvision.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

Author forgot to remove that flag.

Fixes: f7af616c ("net: iosm: infrastructure")
Reported-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The 3rd argument is u32 by function definition while it is __be32
by function declaration.
Signed-off-by: Lijun Pan <lijunp213@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Oleksij Rempel says:

====================
provide cable test support for the ksz886x switch

changes v5:
- drop resume() patch
- add Reviewed-by tags.
- rework dsa_slave_phy_connect() patch

changes v4:
- use fallthrough;
- use EOPNOTSUPP instead of ENOTSUPP
- drop flags variable in dsa_slave_phy_connect patch
- extend description for the "net: phy: micrel: apply resume errat"
  patch
- fix "use consistent alignments" patch

changes v3:
- remove RFC tag

changes v2:
- use generic MII_* defines where possible
- rework phylink validate
- remove phylink get state function
- reorder cabletest patches to make PHY flag patch in the right order
- fix MDI-X detection

This patches provide support for cable testing on the ksz886x switches.
Since it has one special port, we needed to add phylink with validation
and extra quirk for the PHY to signal, that one port will not provide
valid cable testing reports.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch support for cable test for the ksz886x switches and the
ksz8081 PHY.

The patch was tested on a KSZ8873RLL switch with following results:

- port 1:
  - provides invalid values, thus return -ENOTSUPP
    (Errata: DS80000830A: "LinkMD does not work on Port 1",
     http://ww1.microchip.com/downloads/en/DeviceDoc/KSZ8873-Errata-DS80000830A.pdf)

- port 2:
  - can detect distance
  - can detect open on each wire of pair A (wire 1 and 2)
  - can detect open only on one wire of pair B (only wire 3)
  - can detect short between wires of a pair (wires 1 + 2 or 3 + 6)
  - short between pairs is detected as open.
    For example short between wires 2 + 3 is detected as open.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

The current get_phy_flags() is only processed when we connect to a PHY
via a designed phy-handle property via phylink_of_phy_connect(), but if
we fallback on the internal MDIO bus created by a switch and take the
dsa_slave_phy_connect() path then we would not be processing that flag
and using it at PHY connection time.
Suggested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add mapping for LINK_MD register to enable cable testing functionality.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for MDI-X status and configuration
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for MDI-X status and configuration
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch changes the alignments to one space between "#define" and the
macro.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>